Narrow Results

Intensive care unit (ICU) readmission of patients following liver transplantation (LT) is associated with poor outcomes. However, its risk factors remain unclarified. Nowadays, machine learning methods are widely used in many aspects of medical health. This study aims to develop a reliable prognostic model for ICU readmission for post-LT patients using machine learning methods. In this paper, a single center cohort ($N=543$) was studied, of which 5.9% ($N=32$) were readmitted to the ICU during hospitalization for LT. A retrospective review of baseline and perioperative factors possibly related to ICU readmission was performed. Three feature selection techniques were used to detect the best features influencing ICU readmission. Moreover, seven machine learning classifiers were proposed and compared to detect the risk of ICU readmission. Alanine transaminase (ALT) at hospital admission, intraoperative fresh frozen plasma (FFP) and red blood cell (RBC) transfusion, and N-Terminal pro-brain natriuretic peptide (NT-proBNP) after LT were found to be essential features for ICU readmission risk prediction. And the stacking model produced the best performance, identifying patients that were readmitted to the ICU after LT at an accuracy of 97.50%, precision of 96.34%, recall of 96.32%, and F1-score of 96.32%. RBC transfusion is the most crucial feature of the stacking classification model, which produced the best performance with overall accuracy, precision, recall, and F1-score of 88.49%, 88.66%, 76.01%, and 81.84%, respectively.

To explore the risk factors for thyroid nodules and their correlation with diabetes and stroke, the authors conducted a study on 1000 patients with metabolic syndrome (MS). The analysis included variables such as gender, age, familial thyroid disease, salt intake, iodine intake, smoking, alcohol consumption, sleep quality, mental stress, exercise, BMI, blood pressure, diabetes, and baseline nodules. The Apriori algorithm of machine learning was used to derive 12 association rules ($\mathrm{\text{confidence}}\ge 0.5$), and the decision tree algorithm was used to derive 20 effective knowledge rules. The results showed that iodine intake, salt intake, BMI, and advanced age were high-risk factors for thyroid nodules. Exercise, BMI, and age were strongly correlated, while exercise, mental stress, iodine intake, and salt intake showed a strong correlation. Exercise, sleep, smoking, and alcohol consumption influenced mental stress, while age, diseases (diabetes, hypertension, obesity), and lifestyle habits influenced sleep quality. The risk of diabetes and stroke increased in patients with thyroid nodules, and there was a strong correlation among diabetes, stroke, and thyroid nodules.

Objective: The objective of this study was to determine the characteristics that increase the likelihood of acute kidney injury (AKI) in patients with severe community-acquired pneumonia (SCAP) and to create a predictive nomogram for AKI. Methods: This study comprised individuals who received a diagnosis of SCAP over the period from January 01, 2019, to December 31, 2023. The patients were categorized into two groups: AKI and non-AKI. The clinical and demographic characteristics of the patients were extracted from their medical records. An analysis was conducted to compare the rates of survival at 30 and 90 days among various groups. A multivariate analysis was performed to discover the autonomous risk factors linked to SCAP. The nomogram was built based on these parameters. A receiver operating characteristics (ROC) curve study was performed to assess the predictive accuracy of the nomogram, namely by measuring the area under the curve (AUC). Results: Initial screening was conducted on a total of 1218 patients. After excluding 744 individuals who did not meet the exclusion criteria, a total of 474 patients, with an average age of $74.22\pm 15.16$ years and a female representation of 33.3%, were selected for inclusion in this study. The prevalence of AKI in the subjects with SCAP was 47.7%. Out of these instances, 39.8% were categorized as AKI stage 1, 31.0% as AKI stage 2, and 29.2% as AKI stage 3. Those diagnosed with AKI exhibited a significantly higher mortality rate at both the 30-day and 90-day marks in comparison to those who did not have AKI. The independent risk factors for AKI were determined to include age, male gender, chronic renal disease, diabetes, and the utilization of nonsteroidal anti-inflammatory medicines (NSAIDs). In addition, higher levels of baseline serum creatinine and uric acid were identified as risk factors for AKI. The final predictive nomogram achieved an AUC of 0.811, with a 95% confidence interval (CI) ranging from 0.773 to 0.849. Conclusion: Our nomogram can serve as a valuable tool for evaluating the progression of AKI in patients with SCAP.

This study leveraged a large-scale dataset from NHANES 2013–2014 to gain insights into periodontitis pathogenesis and develop predictive tools. After cleaning and preprocessing the data, 15 crucial factors were identified from over 100 potential risk factors and utilized as input features for four machine learning algorithms: support vector machines (SVM), random forest (RF), neural network and XGBoost. The models were evaluated for periodontitis prediction performance through internal validation metrics such as specificity, accuracy, precision, recall and accuracy (area under the curve (AUC)). Notably, education level, household income and smoking status emerged as key risk factors, aligning with medical literature. While SVM and RFs excelled in specificity and accuracy, neural networks surpassed in precision and recall for periodontitis patients. XGBoost offered a balanced performance, making it a versatile choice. The feature importance analysis underscored the profound influence of socioeconomic factors and unhealthy lifestyle habits on periodontal health. This study contributes novel approaches and insights for periodontitis prevention and treatment, demonstrating clinical and societal significance. Future research should focus on optimizing and externally validating the model to enhance its generalizability and accuracy.

All health-related issues exist in a context of extending health expectancy. Behavioral risk factors, diagnostic ”omics,” disparities, insurance, tissue engineering, and climate can shorten life expectancy, but before that, health expectancy. Longer life can bring decades of disability; longer health can mean dying healthy after brief incapacity. Because health precedes other accomplishments, extending average health expectancy into the ninth decade during the $21^{\mathrm{\text{st}}}$ century would have an impact comparable to doubling life expectancy in the $20^{\mathrm{\text{th}}}$ century.

The objective of this research was to build and assess the performance of a prediction model for post-operative recovery status measured by quality of life among individuals experiencing a variety of surgery types. In addition, we assessed the performance of the model for two subgroups (high and moderately consistent wearable device users). Study variables were derived from the electronic health records, questionnaires, and wearable devices of a cohort of individuals with one of 8 surgery types and that were part of the NIH All of Us research program. Through multivariable analysis, high frailty index (OR 1.69, 95% 1.05-7.22, p<0.006), and older age (OR 1.76, 95% 1.55-4.08, p<0.024) were found to be the driving risk factors of poor recovery post-surgery. Our logistic regression model included 15 variables, 5 of which included wearable device data. In wearable use subgroups, the model had better accuracy for high wearable users (81%). Findings demonstrate the potential for models that use wearable measures to assess frailty to inform clinicians of patients at risk for poor surgical outcomes. Our model performed with high accuracy across multiple surgery types and were robust to variable consistency in wearable use.

This PSB 2023 session discusses challenges in clinical implication and application of risk prediction models, which includes but is not limited to: implementation of risk models, responsible use of polygenic risk scores (PGS), and other risk prediction strategies. We focus on the development and use of new, scalable methods for harmonizing and refining risk prediction models by incorporating genetic and non-genetic risk factors, applying new phenotyping strategies, and integrating clinical factors and biomarkers. Lastly, we will discuss innovation in expanding the utility of these prediction models to underrepresented populations. This session focuses on the overarching theme of enabling early diagnosis, and treatment and preventive measures related to complex diseases and comorbidities.